Characterizing and Improving the Robustness of Self-Supervised Learning through Background Augmentations

TL;DR

This paper introduces background augmentations in self-supervised learning to focus models on relevant image content, leading to improved accuracy and robustness across multiple tasks and settings.

Contribution

It proposes simple background augmentation techniques that enhance self-supervised learning by reducing reliance on background cues, achieving state-of-the-art performance.

Findings

+1-2% on ImageNet accuracy

Up to 4.2% improvement in limited-label settings

Enhanced robustness to distribution shifts

Abstract

Recent progress in self-supervised learning has demonstrated promising results in multiple visual tasks. An important ingredient in high-performing self-supervised methods is the use of data augmentation by training models to place different augmented views of the same image nearby in embedding space. However, commonly used augmentation pipelines treat images holistically, ignoring the semantic relevance of parts of an image-e.g. a subject vs. a background-which can lead to the learning of spurious correlations. Our work addresses this problem by investigating a class of simple, yet highly effective "background augmentations", which encourage models to focus on semantically-relevant content by discouraging them from focusing on image backgrounds. Through a systematic investigation, we show that background augmentations lead to substantial improvements in performance across a spectrum of state-of-the-art self-supervised methods (MoCo-v2, BYOL, SwAV) on a variety of tasks, e.g. $\sim$+1-2% gains on ImageNet, enabling performance on par with the supervised baseline. Further, we find the improvement in limited-labels settings is even larger (up to 4.2%). Background augmentations also improve robustness to a number of distribution shifts, including natural adversarial examples, ImageNet-9, adversarial attacks, ImageNet-Renditions. We also make progress in completely unsupervised saliency detection, in the process of generating saliency masks used for background augmentations.